Blame view
mm/workingset.c
21.4 KB
b24413180 License cleanup: ... |
1 |
// SPDX-License-Identifier: GPL-2.0 |
a528910e1 mm: thrash detect... |
2 3 4 5 6 7 8 |
/* * Workingset detection * * Copyright (C) 2013 Red Hat, Inc., Johannes Weiner */ #include <linux/memcontrol.h> |
170b04b7a mm/workingset: pr... |
9 |
#include <linux/mm_inline.h> |
a528910e1 mm: thrash detect... |
10 |
#include <linux/writeback.h> |
3a4f8a0b3 mm: remove shmem_... |
11 |
#include <linux/shmem_fs.h> |
a528910e1 mm: thrash detect... |
12 13 14 15 |
#include <linux/pagemap.h> #include <linux/atomic.h> #include <linux/module.h> #include <linux/swap.h> |
14b468791 mm: workingset: m... |
16 |
#include <linux/dax.h> |
a528910e1 mm: thrash detect... |
17 18 19 20 21 22 |
#include <linux/fs.h> #include <linux/mm.h> /* * Double CLOCK lists * |
1e6b10857 mm, workingset: m... |
23 |
* Per node, two clock lists are maintained for file pages: the |
a528910e1 mm: thrash detect... |
24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 |
* inactive and the active list. Freshly faulted pages start out at * the head of the inactive list and page reclaim scans pages from the * tail. Pages that are accessed multiple times on the inactive list * are promoted to the active list, to protect them from reclaim, * whereas active pages are demoted to the inactive list when the * active list grows too big. * * fault ------------------------+ * | * +--------------+ | +-------------+ * reclaim <- | inactive | <-+-- demotion | active | <--+ * +--------------+ +-------------+ | * | | * +-------------- promotion ------------------+ * * * Access frequency and refault distance * * A workload is thrashing when its pages are frequently used but they * are evicted from the inactive list every time before another access * would have promoted them to the active list. * * In cases where the average access distance between thrashing pages * is bigger than the size of memory there is nothing that can be * done - the thrashing set could never fit into memory under any * circumstance. * * However, the average access distance could be bigger than the * inactive list, yet smaller than the size of memory. In this case, * the set could fit into memory if it weren't for the currently * active pages - which may be used more, hopefully less frequently: * * +-memory available to cache-+ * | | * +-inactive------+-active----+ * a b | c d e f g h i | J K L M N | * +---------------+-----------+ * * It is prohibitively expensive to accurately track access frequency * of pages. But a reasonable approximation can be made to measure * thrashing on the inactive list, after which refaulting pages can be * activated optimistically to compete with the existing active pages. * * Approximating inactive page access frequency - Observations: * * 1. When a page is accessed for the first time, it is added to the * head of the inactive list, slides every existing inactive page * towards the tail by one slot, and pushes the current tail page * out of memory. * * 2. When a page is accessed for the second time, it is promoted to * the active list, shrinking the inactive list by one slot. This * also slides all inactive pages that were faulted into the cache * more recently than the activated page towards the tail of the * inactive list. * * Thus: * * 1. The sum of evictions and activations between any two points in * time indicate the minimum number of inactive pages accessed in * between. * * 2. Moving one inactive page N page slots towards the tail of the * list requires at least N inactive page accesses. * * Combining these: * * 1. When a page is finally evicted from memory, the number of * inactive pages accessed while the page was in cache is at least * the number of page slots on the inactive list. * * 2. In addition, measuring the sum of evictions and activations (E) * at the time of a page's eviction, and comparing it to another * reading (R) at the time the page faults back into memory tells * the minimum number of accesses while the page was not cached. * This is called the refault distance. * * Because the first access of the page was the fault and the second * access the refault, we combine the in-cache distance with the * out-of-cache distance to get the complete minimum access distance * of this page: * * NR_inactive + (R - E) * * And knowing the minimum access distance of a page, we can easily * tell if the page would be able to stay in cache assuming all page * slots in the cache were available: * * NR_inactive + (R - E) <= NR_inactive + NR_active * * which can be further simplified to * * (R - E) <= NR_active * * Put into words, the refault distance (out-of-cache) can be seen as * a deficit in inactive list space (in-cache). If the inactive list * had (R - E) more page slots, the page would not have been evicted * in between accesses, but activated instead. And on a full system, * the only thing eating into inactive list space is active pages. * * |
1899ad18c mm: workingset: t... |
125 |
* Refaulting inactive pages |
a528910e1 mm: thrash detect... |
126 127 128 129 130 131 132 133 134 135 136 137 |
* * All that is known about the active list is that the pages have been * accessed more than once in the past. This means that at any given * time there is actually a good chance that pages on the active list * are no longer in active use. * * So when a refault distance of (R - E) is observed and there are at * least (R - E) active pages, the refaulting page is activated * optimistically in the hope that (R - E) active pages are actually * used less frequently than the refaulting page - or even not used at * all anymore. * |
1899ad18c mm: workingset: t... |
138 139 140 141 |
* That means if inactive cache is refaulting with a suitable refault * distance, we assume the cache workingset is transitioning and put * pressure on the current active list. * |
a528910e1 mm: thrash detect... |
142 143 144 145 146 147 148 |
* If this is wrong and demotion kicks in, the pages which are truly * used more frequently will be reactivated while the less frequently * used once will be evicted from memory. * * But if this is right, the stale pages will be pushed out of memory * and the used pages get to stay in cache. * |
1899ad18c mm: workingset: t... |
149 150 151 152 153 154 155 156 |
* Refaulting active pages * * If on the other hand the refaulting pages have recently been * deactivated, it means that the active list is no longer protecting * actively used cache from reclaim. The cache is NOT transitioning to * a different workingset; the existing workingset is thrashing in the * space allocated to the page cache. * |
a528910e1 mm: thrash detect... |
157 158 159 |
* * Implementation * |
31d8fcac0 mm: workingset: a... |
160 161 |
* For each node's LRU lists, a counter for inactive evictions and * activations is maintained (node->nonresident_age). |
a528910e1 mm: thrash detect... |
162 163 |
* * On eviction, a snapshot of this counter (along with some bits to |
a97e7904c mm: Convert worki... |
164 |
* identify the node) is stored in the now empty page cache |
a528910e1 mm: thrash detect... |
165 166 167 168 169 |
* slot of the evicted page. This is called a shadow entry. * * On cache misses for which there are shadow entries, an eligible * refault distance will immediately activate the refaulting page. */ |
3159f943a xarray: Replace e... |
170 |
#define EVICTION_SHIFT ((BITS_PER_LONG - BITS_PER_XA_VALUE) + \ |
1899ad18c mm: workingset: t... |
171 |
1 + NODES_SHIFT + MEM_CGROUP_ID_SHIFT) |
689c94f03 mm: workingset: #... |
172 |
#define EVICTION_MASK (~0UL >> EVICTION_SHIFT) |
612e44939 mm: workingset: e... |
173 174 |
/* * Eviction timestamps need to be able to cover the full range of |
a97e7904c mm: Convert worki... |
175 |
* actionable refaults. However, bits are tight in the xarray |
612e44939 mm: workingset: e... |
176 177 178 179 180 181 |
* entry, and after storing the identifier for the lruvec there might * not be enough left to represent every single actionable refault. In * that case, we have to sacrifice granularity for distance, and group * evictions into coarser buckets by shaving off lower timestamp bits. */ static unsigned int bucket_order __read_mostly; |
1899ad18c mm: workingset: t... |
182 183 |
static void *pack_shadow(int memcgid, pg_data_t *pgdat, unsigned long eviction, bool workingset) |
a528910e1 mm: thrash detect... |
184 |
{ |
612e44939 mm: workingset: e... |
185 |
eviction >>= bucket_order; |
3159f943a xarray: Replace e... |
186 |
eviction &= EVICTION_MASK; |
23047a96d mm: workingset: p... |
187 |
eviction = (eviction << MEM_CGROUP_ID_SHIFT) | memcgid; |
1e6b10857 mm, workingset: m... |
188 |
eviction = (eviction << NODES_SHIFT) | pgdat->node_id; |
1899ad18c mm: workingset: t... |
189 |
eviction = (eviction << 1) | workingset; |
a528910e1 mm: thrash detect... |
190 |
|
3159f943a xarray: Replace e... |
191 |
return xa_mk_value(eviction); |
a528910e1 mm: thrash detect... |
192 |
} |
1e6b10857 mm, workingset: m... |
193 |
static void unpack_shadow(void *shadow, int *memcgidp, pg_data_t **pgdat, |
1899ad18c mm: workingset: t... |
194 |
unsigned long *evictionp, bool *workingsetp) |
a528910e1 mm: thrash detect... |
195 |
{ |
3159f943a xarray: Replace e... |
196 |
unsigned long entry = xa_to_value(shadow); |
1e6b10857 mm, workingset: m... |
197 |
int memcgid, nid; |
1899ad18c mm: workingset: t... |
198 |
bool workingset; |
a528910e1 mm: thrash detect... |
199 |
|
1899ad18c mm: workingset: t... |
200 201 |
workingset = entry & 1; entry >>= 1; |
a528910e1 mm: thrash detect... |
202 203 |
nid = entry & ((1UL << NODES_SHIFT) - 1); entry >>= NODES_SHIFT; |
23047a96d mm: workingset: p... |
204 205 |
memcgid = entry & ((1UL << MEM_CGROUP_ID_SHIFT) - 1); entry >>= MEM_CGROUP_ID_SHIFT; |
a528910e1 mm: thrash detect... |
206 |
|
23047a96d mm: workingset: p... |
207 |
*memcgidp = memcgid; |
1e6b10857 mm, workingset: m... |
208 |
*pgdat = NODE_DATA(nid); |
612e44939 mm: workingset: e... |
209 |
*evictionp = entry << bucket_order; |
1899ad18c mm: workingset: t... |
210 |
*workingsetp = workingset; |
a528910e1 mm: thrash detect... |
211 |
} |
31d8fcac0 mm: workingset: a... |
212 213 |
/** * workingset_age_nonresident - age non-resident entries as LRU ages |
e755f4af0 mm/workingset.c: ... |
214 |
* @lruvec: the lruvec that was aged |
31d8fcac0 mm: workingset: a... |
215 216 217 218 219 220 221 222 |
* @nr_pages: the number of pages to count * * As in-memory pages are aged, non-resident pages need to be aged as * well, in order for the refault distances later on to be comparable * to the in-memory dimensions. This function allows reclaim and LRU * operations to drive the non-resident aging along in parallel. */ void workingset_age_nonresident(struct lruvec *lruvec, unsigned long nr_pages) |
b910718a9 mm: vmscan: detec... |
223 224 225 226 227 228 229 230 231 232 233 234 235 |
{ /* * Reclaiming a cgroup means reclaiming all its children in a * round-robin fashion. That means that each cgroup has an LRU * order that is composed of the LRU orders of its child * cgroups; and every page has an LRU position not just in the * cgroup that owns it, but in all of that group's ancestors. * * So when the physical inactive list of a leaf cgroup ages, * the virtual inactive lists of all its parents, including * the root cgroup's, age as well. */ do { |
31d8fcac0 mm: workingset: a... |
236 237 |
atomic_long_add(nr_pages, &lruvec->nonresident_age); } while ((lruvec = parent_lruvec(lruvec))); |
b910718a9 mm: vmscan: detec... |
238 |
} |
a528910e1 mm: thrash detect... |
239 240 |
/** * workingset_eviction - note the eviction of a page from memory |
b910718a9 mm: vmscan: detec... |
241 |
* @target_memcg: the cgroup that is causing the reclaim |
a528910e1 mm: thrash detect... |
242 243 |
* @page: the page being evicted * |
a7ca12f9d mm/workingset: re... |
244 |
* Returns a shadow entry to be stored in @page->mapping->i_pages in place |
a528910e1 mm: thrash detect... |
245 246 |
* of the evicted @page so that a later refault can be detected. */ |
b910718a9 mm: vmscan: detec... |
247 |
void *workingset_eviction(struct page *page, struct mem_cgroup *target_memcg) |
a528910e1 mm: thrash detect... |
248 |
{ |
1e6b10857 mm, workingset: m... |
249 |
struct pglist_data *pgdat = page_pgdat(page); |
a528910e1 mm: thrash detect... |
250 |
unsigned long eviction; |
23047a96d mm: workingset: p... |
251 |
struct lruvec *lruvec; |
b910718a9 mm: vmscan: detec... |
252 |
int memcgid; |
a528910e1 mm: thrash detect... |
253 |
|
23047a96d mm: workingset: p... |
254 255 256 257 |
/* Page is fully exclusive and pins page->mem_cgroup */ VM_BUG_ON_PAGE(PageLRU(page), page); VM_BUG_ON_PAGE(page_count(page), page); VM_BUG_ON_PAGE(!PageLocked(page), page); |
b910718a9 mm: vmscan: detec... |
258 |
lruvec = mem_cgroup_lruvec(target_memcg, pgdat); |
6c357848b mm: replace hpage... |
259 |
workingset_age_nonresident(lruvec, thp_nr_pages(page)); |
b910718a9 mm: vmscan: detec... |
260 261 |
/* XXX: target_memcg can be NULL, go through lruvec */ memcgid = mem_cgroup_id(lruvec_memcg(lruvec)); |
31d8fcac0 mm: workingset: a... |
262 |
eviction = atomic_long_read(&lruvec->nonresident_age); |
1899ad18c mm: workingset: t... |
263 |
return pack_shadow(memcgid, pgdat, eviction, PageWorkingset(page)); |
a528910e1 mm: thrash detect... |
264 265 266 267 |
} /** * workingset_refault - evaluate the refault of a previously evicted page |
1899ad18c mm: workingset: t... |
268 |
* @page: the freshly allocated replacement page |
a528910e1 mm: thrash detect... |
269 270 271 |
* @shadow: shadow entry of the evicted page * * Calculates and evaluates the refault distance of the previously |
b910718a9 mm: vmscan: detec... |
272 273 |
* evicted page in the context of the node and the memcg whose memory * pressure caused the eviction. |
a528910e1 mm: thrash detect... |
274 |
*/ |
1899ad18c mm: workingset: t... |
275 |
void workingset_refault(struct page *page, void *shadow) |
a528910e1 mm: thrash detect... |
276 |
{ |
170b04b7a mm/workingset: pr... |
277 |
bool file = page_is_file_lru(page); |
b910718a9 mm: vmscan: detec... |
278 279 |
struct mem_cgroup *eviction_memcg; struct lruvec *eviction_lruvec; |
a528910e1 mm: thrash detect... |
280 |
unsigned long refault_distance; |
34e58cac6 mm: workingset: l... |
281 |
unsigned long workingset_size; |
1899ad18c mm: workingset: t... |
282 |
struct pglist_data *pgdat; |
23047a96d mm: workingset: p... |
283 |
struct mem_cgroup *memcg; |
162453bfb mm: workingset: s... |
284 |
unsigned long eviction; |
23047a96d mm: workingset: p... |
285 |
struct lruvec *lruvec; |
162453bfb mm: workingset: s... |
286 |
unsigned long refault; |
1899ad18c mm: workingset: t... |
287 |
bool workingset; |
23047a96d mm: workingset: p... |
288 |
int memcgid; |
a528910e1 mm: thrash detect... |
289 |
|
1899ad18c mm: workingset: t... |
290 |
unpack_shadow(shadow, &memcgid, &pgdat, &eviction, &workingset); |
162453bfb mm: workingset: s... |
291 |
|
23047a96d mm: workingset: p... |
292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 |
rcu_read_lock(); /* * Look up the memcg associated with the stored ID. It might * have been deleted since the page's eviction. * * Note that in rare events the ID could have been recycled * for a new cgroup that refaults a shared page. This is * impossible to tell from the available data. However, this * should be a rare and limited disturbance, and activations * are always speculative anyway. Ultimately, it's the aging * algorithm's job to shake out the minimum access frequency * for the active cache. * * XXX: On !CONFIG_MEMCG, this will always return NULL; it * would be better if the root_mem_cgroup existed in all * configurations instead. */ |
b910718a9 mm: vmscan: detec... |
309 310 |
eviction_memcg = mem_cgroup_from_id(memcgid); if (!mem_cgroup_disabled() && !eviction_memcg) |
1899ad18c mm: workingset: t... |
311 |
goto out; |
b910718a9 mm: vmscan: detec... |
312 |
eviction_lruvec = mem_cgroup_lruvec(eviction_memcg, pgdat); |
31d8fcac0 mm: workingset: a... |
313 |
refault = atomic_long_read(&eviction_lruvec->nonresident_age); |
162453bfb mm: workingset: s... |
314 315 |
/* |
1899ad18c mm: workingset: t... |
316 |
* Calculate the refault distance |
162453bfb mm: workingset: s... |
317 |
* |
1899ad18c mm: workingset: t... |
318 |
* The unsigned subtraction here gives an accurate distance |
31d8fcac0 mm: workingset: a... |
319 |
* across nonresident_age overflows in most cases. There is a |
1899ad18c mm: workingset: t... |
320 321 322 |
* special case: usually, shadow entries have a short lifetime * and are either refaulted or reclaimed along with the inode * before they get too old. But it is not impossible for the |
31d8fcac0 mm: workingset: a... |
323 324 325 326 |
* nonresident_age to lap a shadow entry in the field, which * can then result in a false small refault distance, leading * to a false activation should this old entry actually * refault again. However, earlier kernels used to deactivate |
1899ad18c mm: workingset: t... |
327 328 329 |
* unconditionally with *every* reclaim invocation for the * longest time, so the occasional inappropriate activation * leading to pressure on the active list is not a problem. |
162453bfb mm: workingset: s... |
330 331 |
*/ refault_distance = (refault - eviction) & EVICTION_MASK; |
b910718a9 mm: vmscan: detec... |
332 333 334 335 336 337 338 339 340 341 |
/* * The activation decision for this page is made at the level * where the eviction occurred, as that is where the LRU order * during page reclaim is being determined. * * However, the cgroup that will own the page is the one that * is actually experiencing the refault event. */ memcg = page_memcg(page); lruvec = mem_cgroup_lruvec(memcg, pgdat); |
170b04b7a mm/workingset: pr... |
342 |
inc_lruvec_state(lruvec, WORKINGSET_REFAULT_BASE + file); |
a528910e1 mm: thrash detect... |
343 |
|
1899ad18c mm: workingset: t... |
344 345 |
/* * Compare the distance to the existing workingset size. We |
34e58cac6 mm: workingset: l... |
346 |
* don't activate pages that couldn't stay resident even if |
aae466b00 mm/swap: implemen... |
347 348 349 |
* all the memory was available to the workingset. Whether * workingset competition needs to consider anon or not depends * on having swap. |
1899ad18c mm: workingset: t... |
350 |
*/ |
34e58cac6 mm: workingset: l... |
351 |
workingset_size = lruvec_page_state(eviction_lruvec, NR_ACTIVE_FILE); |
aae466b00 mm/swap: implemen... |
352 |
if (!file) { |
34e58cac6 mm: workingset: l... |
353 |
workingset_size += lruvec_page_state(eviction_lruvec, |
aae466b00 mm/swap: implemen... |
354 355 356 |
NR_INACTIVE_FILE); } if (mem_cgroup_get_nr_swap_pages(memcg) > 0) { |
34e58cac6 mm: workingset: l... |
357 358 |
workingset_size += lruvec_page_state(eviction_lruvec, NR_ACTIVE_ANON); |
aae466b00 mm/swap: implemen... |
359 360 361 362 |
if (file) { workingset_size += lruvec_page_state(eviction_lruvec, NR_INACTIVE_ANON); } |
34e58cac6 mm: workingset: l... |
363 364 |
} if (refault_distance > workingset_size) |
1899ad18c mm: workingset: t... |
365 366 367 |
goto out; SetPageActive(page); |
6c357848b mm: replace hpage... |
368 |
workingset_age_nonresident(lruvec, thp_nr_pages(page)); |
170b04b7a mm/workingset: pr... |
369 |
inc_lruvec_state(lruvec, WORKINGSET_ACTIVATE_BASE + file); |
1899ad18c mm: workingset: t... |
370 371 372 373 |
/* Page was active prior to eviction */ if (workingset) { SetPageWorkingset(page); |
314b57fb0 mm: balance LRU l... |
374 375 |
/* XXX: Move to lru_cache_add() when it supports new vs putback */ spin_lock_irq(&page_pgdat(page)->lru_lock); |
96f8bf4fb mm: vmscan: recla... |
376 |
lru_note_cost_page(page); |
314b57fb0 mm: balance LRU l... |
377 |
spin_unlock_irq(&page_pgdat(page)->lru_lock); |
170b04b7a mm/workingset: pr... |
378 |
inc_lruvec_state(lruvec, WORKINGSET_RESTORE_BASE + file); |
a528910e1 mm: thrash detect... |
379 |
} |
1899ad18c mm: workingset: t... |
380 |
out: |
2a2e48854 mm: vmscan: fix I... |
381 |
rcu_read_unlock(); |
a528910e1 mm: thrash detect... |
382 383 384 385 386 387 388 389 |
} /** * workingset_activation - note a page activation * @page: page that is being activated */ void workingset_activation(struct page *page) { |
55779ec75 mm: fix vm-scalab... |
390 |
struct mem_cgroup *memcg; |
31d8fcac0 mm: workingset: a... |
391 |
struct lruvec *lruvec; |
23047a96d mm: workingset: p... |
392 |
|
55779ec75 mm: fix vm-scalab... |
393 |
rcu_read_lock(); |
23047a96d mm: workingset: p... |
394 395 396 397 398 399 400 |
/* * Filter non-memcg pages here, e.g. unmap can call * mark_page_accessed() on VDSO pages. * * XXX: See workingset_refault() - this should return * root_mem_cgroup even for !CONFIG_MEMCG. */ |
55779ec75 mm: fix vm-scalab... |
401 402 |
memcg = page_memcg_rcu(page); if (!mem_cgroup_disabled() && !memcg) |
23047a96d mm: workingset: p... |
403 |
goto out; |
31d8fcac0 mm: workingset: a... |
404 |
lruvec = mem_cgroup_page_lruvec(page, page_pgdat(page)); |
6c357848b mm: replace hpage... |
405 |
workingset_age_nonresident(lruvec, thp_nr_pages(page)); |
23047a96d mm: workingset: p... |
406 |
out: |
55779ec75 mm: fix vm-scalab... |
407 |
rcu_read_unlock(); |
a528910e1 mm: thrash detect... |
408 |
} |
449dd6984 mm: keep page cac... |
409 410 411 412 413 414 415 416 417 418 419 420 |
/* * Shadow entries reflect the share of the working set that does not * fit into memory, so their number depends on the access pattern of * the workload. In most cases, they will refault or get reclaimed * along with the inode, but a (malicious) workload that streams * through files with a total size several times that of available * memory, while preventing the inodes from being reclaimed, can * create excessive amounts of shadow nodes. To keep a lid on this, * track shadow nodes and reclaim them when they grow way past the * point where they would still be useful. */ |
14b468791 mm: workingset: m... |
421 |
static struct list_lru shadow_nodes; |
a97e7904c mm: Convert worki... |
422 |
void workingset_update_node(struct xa_node *node) |
14b468791 mm: workingset: m... |
423 |
{ |
14b468791 mm: workingset: m... |
424 425 426 427 428 429 |
/* * Track non-empty nodes that contain only shadow entries; * unlink those that contain pages or are being freed. * * Avoid acquiring the list_lru lock when the nodes are * already where they should be. The list_empty() test is safe |
b93b01631 page cache: use x... |
430 |
* as node->private_list is protected by the i_pages lock. |
14b468791 mm: workingset: m... |
431 |
*/ |
68d48e6a2 mm: workingset: a... |
432 |
VM_WARN_ON_ONCE(!irqs_disabled()); /* For __inc_lruvec_page_state */ |
01959dfe7 xarray: Define st... |
433 |
if (node->count && node->count == node->nr_values) { |
68d48e6a2 mm: workingset: a... |
434 |
if (list_empty(&node->private_list)) { |
14b468791 mm: workingset: m... |
435 |
list_lru_add(&shadow_nodes, &node->private_list); |
ec9f02384 mm: workingset: f... |
436 |
__inc_lruvec_slab_state(node, WORKINGSET_NODES); |
68d48e6a2 mm: workingset: a... |
437 |
} |
14b468791 mm: workingset: m... |
438 |
} else { |
68d48e6a2 mm: workingset: a... |
439 |
if (!list_empty(&node->private_list)) { |
14b468791 mm: workingset: m... |
440 |
list_lru_del(&shadow_nodes, &node->private_list); |
ec9f02384 mm: workingset: f... |
441 |
__dec_lruvec_slab_state(node, WORKINGSET_NODES); |
68d48e6a2 mm: workingset: a... |
442 |
} |
14b468791 mm: workingset: m... |
443 444 |
} } |
449dd6984 mm: keep page cac... |
445 446 447 448 |
static unsigned long count_shadow_nodes(struct shrinker *shrinker, struct shrink_control *sc) { |
449dd6984 mm: keep page cac... |
449 |
unsigned long max_nodes; |
14b468791 mm: workingset: m... |
450 |
unsigned long nodes; |
95f9ab2d5 mm: workingset: d... |
451 |
unsigned long pages; |
449dd6984 mm: keep page cac... |
452 |
|
14b468791 mm: workingset: m... |
453 |
nodes = list_lru_shrink_count(&shadow_nodes, sc); |
449dd6984 mm: keep page cac... |
454 |
|
449dd6984 mm: keep page cac... |
455 |
/* |
a97e7904c mm: Convert worki... |
456 |
* Approximate a reasonable limit for the nodes |
b53889987 mm: workingset: u... |
457 458 459 460 461 462 463 464 465 466 467 468 469 |
* containing shadow entries. We don't need to keep more * shadow entries than possible pages on the active list, * since refault distances bigger than that are dismissed. * * The size of the active list converges toward 100% of * overall page cache as memory grows, with only a tiny * inactive list. Assume the total cache size for that. * * Nodes might be sparsely populated, with only one shadow * entry in the extreme case. Obviously, we cannot keep one * node for every eligible shadow entry, so compromise on a * worst-case density of 1/8th. Below that, not all eligible * refaults can be detected anymore. |
449dd6984 mm: keep page cac... |
470 |
* |
a97e7904c mm: Convert worki... |
471 |
* On 64-bit with 7 xa_nodes per page and 64 slots |
449dd6984 mm: keep page cac... |
472 |
* each, this will reclaim shadow entries when they consume |
b53889987 mm: workingset: u... |
473 |
* ~1.8% of available memory: |
449dd6984 mm: keep page cac... |
474 |
* |
a97e7904c mm: Convert worki... |
475 |
* PAGE_SIZE / xa_nodes / node_entries * 8 / PAGE_SIZE |
449dd6984 mm: keep page cac... |
476 |
*/ |
95f9ab2d5 mm: workingset: d... |
477 |
#ifdef CONFIG_MEMCG |
b53889987 mm: workingset: u... |
478 |
if (sc->memcg) { |
95f9ab2d5 mm: workingset: d... |
479 |
struct lruvec *lruvec; |
2b487e59f mm: memcontrol: p... |
480 |
int i; |
95f9ab2d5 mm: workingset: d... |
481 |
|
867e5e1de mm: clean up and ... |
482 |
lruvec = mem_cgroup_lruvec(sc->memcg, NODE_DATA(sc->nid)); |
2b487e59f mm: memcontrol: p... |
483 |
for (pages = 0, i = 0; i < NR_LRU_LISTS; i++) |
205b20cc5 mm: memcontrol: m... |
484 485 |
pages += lruvec_page_state_local(lruvec, NR_LRU_BASE + i); |
d42f3245c mm: memcg: conver... |
486 487 488 489 |
pages += lruvec_page_state_local( lruvec, NR_SLAB_RECLAIMABLE_B) >> PAGE_SHIFT; pages += lruvec_page_state_local( lruvec, NR_SLAB_UNRECLAIMABLE_B) >> PAGE_SHIFT; |
95f9ab2d5 mm: workingset: d... |
490 491 492 |
} else #endif pages = node_present_pages(sc->nid); |
dad4f140e Merge branch 'xar... |
493 |
max_nodes = pages >> (XA_CHUNK_SHIFT - 3); |
449dd6984 mm: keep page cac... |
494 |
|
9b996468c mm: add SHRINK_EM... |
495 496 |
if (!nodes) return SHRINK_EMPTY; |
14b468791 mm: workingset: m... |
497 |
if (nodes <= max_nodes) |
449dd6984 mm: keep page cac... |
498 |
return 0; |
14b468791 mm: workingset: m... |
499 |
return nodes - max_nodes; |
449dd6984 mm: keep page cac... |
500 501 502 |
} static enum lru_status shadow_lru_isolate(struct list_head *item, |
3f97b1632 list_lru: add hel... |
503 |
struct list_lru_one *lru, |
449dd6984 mm: keep page cac... |
504 |
spinlock_t *lru_lock, |
a97e7904c mm: Convert worki... |
505 |
void *arg) __must_hold(lru_lock) |
449dd6984 mm: keep page cac... |
506 |
{ |
a97e7904c mm: Convert worki... |
507 |
struct xa_node *node = container_of(item, struct xa_node, private_list); |
449dd6984 mm: keep page cac... |
508 |
struct address_space *mapping; |
449dd6984 mm: keep page cac... |
509 510 511 |
int ret; /* |
f82cd2f0b XArray: Add priva... |
512 |
* Page cache insertions and deletions synchronously maintain |
b93b01631 page cache: use x... |
513 |
* the shadow node LRU under the i_pages lock and the |
449dd6984 mm: keep page cac... |
514 515 |
* lru_lock. Because the page cache tree is emptied before * the inode can be destroyed, holding the lru_lock pins any |
a97e7904c mm: Convert worki... |
516 |
* address_space that has nodes on the LRU. |
449dd6984 mm: keep page cac... |
517 |
* |
b93b01631 page cache: use x... |
518 |
* We can then safely transition to the i_pages lock to |
449dd6984 mm: keep page cac... |
519 520 521 |
* pin only the address_space of the particular node we want * to reclaim, take the node off-LRU, and drop the lru_lock. */ |
01959dfe7 xarray: Define st... |
522 |
mapping = container_of(node->array, struct address_space, i_pages); |
449dd6984 mm: keep page cac... |
523 524 |
/* Coming from the list, invert the lock order */ |
b93b01631 page cache: use x... |
525 |
if (!xa_trylock(&mapping->i_pages)) { |
6ca342d02 mm: workingset: m... |
526 |
spin_unlock_irq(lru_lock); |
449dd6984 mm: keep page cac... |
527 528 529 |
ret = LRU_RETRY; goto out; } |
3f97b1632 list_lru: add hel... |
530 |
list_lru_isolate(lru, item); |
ec9f02384 mm: workingset: f... |
531 |
__dec_lruvec_slab_state(node, WORKINGSET_NODES); |
68d48e6a2 mm: workingset: a... |
532 |
|
449dd6984 mm: keep page cac... |
533 534 535 536 537 538 539 |
spin_unlock(lru_lock); /* * The nodes should only contain one or more shadow entries, * no pages, so we expect to be able to remove them all and * delete and free the empty node afterwards. */ |
01959dfe7 xarray: Define st... |
540 |
if (WARN_ON_ONCE(!node->nr_values)) |
b936887e8 mm: workingset: t... |
541 |
goto out_invalid; |
01959dfe7 xarray: Define st... |
542 |
if (WARN_ON_ONCE(node->count != node->nr_values)) |
b936887e8 mm: workingset: t... |
543 |
goto out_invalid; |
a97e7904c mm: Convert worki... |
544 |
mapping->nrexceptional -= node->nr_values; |
f82cd2f0b XArray: Add priva... |
545 |
xa_delete_node(node, workingset_update_node); |
ec9f02384 mm: workingset: f... |
546 |
__inc_lruvec_slab_state(node, WORKINGSET_NODERECLAIM); |
449dd6984 mm: keep page cac... |
547 |
|
b936887e8 mm: workingset: t... |
548 |
out_invalid: |
6ca342d02 mm: workingset: m... |
549 |
xa_unlock_irq(&mapping->i_pages); |
449dd6984 mm: keep page cac... |
550 551 |
ret = LRU_REMOVED_RETRY; out: |
449dd6984 mm: keep page cac... |
552 |
cond_resched(); |
6ca342d02 mm: workingset: m... |
553 |
spin_lock_irq(lru_lock); |
449dd6984 mm: keep page cac... |
554 555 556 557 558 559 |
return ret; } static unsigned long scan_shadow_nodes(struct shrinker *shrinker, struct shrink_control *sc) { |
b93b01631 page cache: use x... |
560 |
/* list_lru lock nests inside the IRQ-safe i_pages lock */ |
6b51e8819 mm/list_lru: intr... |
561 562 |
return list_lru_shrink_walk_irq(&shadow_nodes, sc, shadow_lru_isolate, NULL); |
449dd6984 mm: keep page cac... |
563 564 565 566 567 |
} static struct shrinker workingset_shadow_shrinker = { .count_objects = count_shadow_nodes, .scan_objects = scan_shadow_nodes, |
4b85afbda mm: zero-seek shr... |
568 |
.seeks = 0, /* ->count reports only fully expendable nodes */ |
0a6b76dd2 mm: workingset: m... |
569 |
.flags = SHRINKER_NUMA_AWARE | SHRINKER_MEMCG_AWARE, |
449dd6984 mm: keep page cac... |
570 571 572 573 |
}; /* * Our list_lru->lock is IRQ-safe as it nests inside the IRQ-safe |
b93b01631 page cache: use x... |
574 |
* i_pages lock. |
449dd6984 mm: keep page cac... |
575 576 577 578 579 |
*/ static struct lock_class_key shadow_nodes_key; static int __init workingset_init(void) { |
612e44939 mm: workingset: e... |
580 581 |
unsigned int timestamp_bits; unsigned int max_order; |
449dd6984 mm: keep page cac... |
582 |
int ret; |
612e44939 mm: workingset: e... |
583 584 585 586 587 588 589 590 591 |
BUILD_BUG_ON(BITS_PER_LONG < EVICTION_SHIFT); /* * Calculate the eviction bucket size to cover the longest * actionable refault distance, which is currently half of * memory (totalram_pages/2). However, memory hotplug may add * some more pages at runtime, so keep working with up to * double the initial memory by using totalram_pages as-is. */ timestamp_bits = BITS_PER_LONG - EVICTION_SHIFT; |
ca79b0c21 mm: convert total... |
592 |
max_order = fls_long(totalram_pages() - 1); |
612e44939 mm: workingset: e... |
593 594 |
if (max_order > timestamp_bits) bucket_order = max_order - timestamp_bits; |
d3d36c4b5 mm: workingset: p... |
595 596 |
pr_info("workingset: timestamp_bits=%d max_order=%d bucket_order=%u ", |
612e44939 mm: workingset: e... |
597 |
timestamp_bits, max_order, bucket_order); |
39887653a mm/workingset.c: ... |
598 |
ret = prealloc_shrinker(&workingset_shadow_shrinker); |
449dd6984 mm: keep page cac... |
599 600 |
if (ret) goto err; |
c92e8e10c fs: propagate shr... |
601 602 |
ret = __list_lru_init(&shadow_nodes, true, &shadow_nodes_key, &workingset_shadow_shrinker); |
449dd6984 mm: keep page cac... |
603 604 |
if (ret) goto err_list_lru; |
39887653a mm/workingset.c: ... |
605 |
register_shrinker_prepared(&workingset_shadow_shrinker); |
449dd6984 mm: keep page cac... |
606 607 |
return 0; err_list_lru: |
39887653a mm/workingset.c: ... |
608 |
free_prealloced_shrinker(&workingset_shadow_shrinker); |
449dd6984 mm: keep page cac... |
609 610 611 612 |
err: return ret; } module_init(workingset_init); |